Extremely Different Results in Cluster vs. PC

[skamalzare]

Hi,

I have a simple 3D model. It applies a ground motion to base of a reinforced soil profile and records the accelerations at its top surface. I run the code on my PC and results are just fine. Then I copy-paste the code to VT-Cluster server (linux machine) and run it there. The results are in order of 1e-14 different. It seems like that input motion somewhere gets significantly small. The accelerations still have the form of the input motion but they are ~1e-14 smaller. I would greatly appreciate if you could tell me where is my problem, and possibly, how I can fix it.

Thanks,
Soheil

[fmk]

record the displacements at the base to see what they are.

[skamalzare]

Dear Frank,

Thanks for the comment. The problem was that VT-Cluster server had OpenSees/2.4.5 while I am using the latest version.

I checked the displacements and noticed that I am assigning fix boundaries to the base two times during gravity analyses (once before the elastic state and once before the plastic state). But before dynamic analyses, I was removing them with "remove SP" command. Apparently, OpenSees 2.4.5 was not completely removing all the fixities while OpenSees 2.5.0 was completely removing them.

Anyways, I asked them to update their OpenSees and come-up with better BCs in my code. So it is fixed now.

Thanks a lot for your help,
Soheil